Twenty-eight radiologists read the same 200 screening mammograms, representing a total of 5587 readings (13 lost readings). Based on the interpretations of these radiologists, the false-positive rate was 36% (1728 of 4750) and the false-negative rate was 16% (132 of 837). Therefore, the average sensitivity was 84% (705 of 837) (range, 63-97%) and the average specificity was 64% (3022 of 4750) (range, 34-85%) (Table 1).
Table 1
The seven radiologists not routinely interpreting mammograms showed an average sensitivity of 84% (177 of 210) (range, 63-97%) and an average specificity of 56% (671 of 1189) (range, 34-69%) while the 21 routine readers showed the similar average sensitivity of 85% (531 of 627) (range, 63-97%) (p=.999) but a higher specificity of 66% (2351 of 3561) (range, 51-85%) (p<.001). The global measure of accuracy revealed that 61% (848 of 1399) of readings were correctly classified in the group of radiologists not routinely interpreting mammograms compared with 69% (2882 of 4188) in the groups of routine readers (p<.001) (data not shown).
The 21 most experienced radiologists had a mean age of 47 years (range, 40-60 years), had 12 years’ experience of reading mammograms (range, 4-22 years), had read an average of 5773 mammograms in the year prior to participating in the study (range, 1890-13230 mammograms), and spent an average 56% of their working hours on breast disease (range, 15-100%). Eighty-one percent (17 of 21) of the radiologists routinely consulted colleagues and 86% (18 of 21) routinely obtained feedback on cases for which they recommended further workup (data not shown). Given the characteristics defining the group of radiologists not routinely interpreting mammograms, experience-related variables were not evaluated in these seven radiologists.
Figure 1 shows radiologists’ sensitivity, specificity and accuracy according to the distinct experience-related variables. No significant differences in sensitivity were observed between the group of radiologists not routinely interpreting mammograms and the less experienced radiologists among the group of routine readers according to any of the factors under study, i.e. reader volume, obtaining feedback, percentage of time spent reading mammograms in clinical practice, consultations with colleagues, or age. In contrast, in the group of radiologists not routinely interpreting mammograms, specificity was significantly lower in all the above-mentioned variables (all p<.001). For the variable of accuracy, and except for the variables of feedback (p=.254) and focus on mammogram reading (p=.066), radiologists not routinely interpreting mammograms were significantly less accurate (all p<.05).
Figure 1
When routine readers only were considered, those spending more than 25% of their working day on mammogram reading showed higher sensitivity (86% vs 78%, p=.019) but lower specificity (65% vs 70%, p<.0001). In contrast, those consulting with colleagues showed lower sensitivity (83% vs 90%, p=.036) and higher specificity (67% vs 62%, p<.0001), and those aged more than 45 years old also showed lower sensitivity (81% vs 90%, p=.007) and higher specificity (68% vs 58%, p<.0001).
The multivariate model was used to evaluate the 21 radiologists routinely interpreting mammograms. The only measure showing statistically significant differences was specificity, which was higher in radiologists obtaining feedback (OR=1.37, 95% CI=[1.03; 1.85]), in those devoting less than 25% of their working hours to mammogram reading (OR=1.49, 95% CI=[1.18; 1.89]), and in those aged more than 45 years (OR=1.33, 95% CI=[1.12; 1.59]). The remaining variables (annual reading volume, years of experience in reading mammograms and consultation with colleagues) showed no influence on sensitivity, specificity or accuracy (Table 2).
Table 2